ESXi Host Unresponsiveness and Admission Failures due to Excessive Memory Reservations

Products

• VMware vSphere ESXi

Issue/Introduction

• An ESXi host becomes unmanageable from vCenter and SSH.
• VMs may continue to run but cannot be managed, and new VMs fail to power on.
• ESXi host's logs show widespread errors such as:
  • vob.visorfs.ramdisk.allocfail
  • Admission failure events related to visorfsPager and hardware drivers
  • PANIC: Unexpected signal: 11 in management libraries.

Environment

VMware ESXi

Cause

• The host is experiencing severe memory starvation because the total memory reserved for Virtual Machines exceeds the safe operating threshold for ESXi.
  E.g.:
  • A host has 512 GB of physical RAM
  • 488 GB (95.3%) is reserved by 4 VMs (all VMs have 100% memory reservation).
  • This leaves only ~24 GB for the ESXi kernel, drivers, and management agents, which is below the recommended 6% overhead required for stable operations on high-memory hosts.

Resolution

This is expected behaviour.
To resolve and prevent this state:

1. Reboot the affected host to clear the memory starvation state and recover the ramdisk.
2. Review VM Memory Reservations to ensure that the total of all VM memory reservations running on the ESXi host does not exceed 94% of physical RAM.
3. Follow Overhead Guidelines: Refer to the following table for minimum recommended overhead:
   • 2-socket hosts (512GB - 1TB): Minimum 32GB - 64GB overhead.
   • 2-socket hosts (2TB+): Minimum 128GB overhead.

Additional Information

Estimated memory requirements for ESXi hosts running with very large memory reservations by single or multiple virtual machines